Robust Identification of Autoregressive Moving Average Models

SUMMARY We introduce a class of robust estimates for the partial autocorrelation function of a univariate stationary time series and show that it is possible to produce an estimate of the autocorrelation function from the estimated partial autocorrelation coefficients. These statistics seem suitable for the preliminary identification of the order, p and q of an ARMA (p, q) model when the observed series contains a few aberrant observations.